Methods and systems for vasculature characterization using ultrasound

Abstract

There is provided a method and a system for vasculature characterization using ultrasound. Ultrasound data is continuously acquired from a body comprising a vasculature with one or more ultrasound contrast agents flowing therethrough. The ultrasound data is filtered to separate background signals from reflected ultrasound signals generated by the one or more ultrasound contrast agents. An entire trajectory of each of the one or more ultrasound contrast agents through the vasculature is tracked based on the reflected ultrasound signals. At least one displacement characteristic of each of the one or more ultrasound contrast agents throughout the entire trajectory over time is determined based on the tracking. One or more vessels present within the vasculature are individually identified based on the at least one displacement characteristic. At least one image of the vasculature including the one or more vessels may be generated and output.

Description

METHODS AND SYSTEMS FOR VASCULATURE CHARACTERIZATION USING ULTRASOUNDCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] The present application claims the benefit of United States Provisional Patent Application No. 63 / 651 ,545 filed on May 24, 2024, the contents of which are hereby incorporated by reference.FIELD

[0002] The improvements generally relate to the field of ultrasound, and more particularly to characterization of vasculature using ultrasound.BACKGROUND

[0003] The microvascular system within organs, such as the brain, comprising a complex network of blood vessels including arteries, arterioles, capillaries, venules, and veins, plays a fundamental role in maintaining the homeostasis and integrity of living systems, such as the central nervous system. Neuronal health is intimately linked to the proper function of the microvasculature and emerging evidence suggests that disruptions in microvascular function are associated with the onset, maintenance, and progression of neurodegenerative conditions. Notably, detection of disruptions at the capillary level, where oxygen exchange occurs and red blood cell velocity is at its lowest, is of particular importance since anomalies in microvascular flow dynamics, such as turbulent flow, endothelial damage, or intermittent red blood cell stalling, have been correlated with cognitive decline and worsening neurodegeneration. Moreover, disruptions at the single capillary level can cascade downstream, impacting the entire microvasculature and neural networks. However, there is currently no commercially available in vivo imaging technique capable of precisely assessing single capillary function throughout whole organs, such as the brain.

[0004] Therefore, improvements are needed.SUMMARY

[0005] In accordance with one aspect, there is provided a method for vasculature characterization using ultrasound. The method comprises continuously acquiring ultrasound data from a body comprising a vasculature with one or more ultrasoundcontrast agents flowing therethrough, filtering the ultrasound data to separate background signals from reflected ultrasound signals generated by the one or more ultrasound contrast agents, tracking, based on the reflected ultrasound signals, an entire trajectory of each of the one or more ultrasound contrast agents through the vasculature, determining, based on the tracking, at least one displacement characteristic of each of the one or more ultrasound contrast agents throughout the entire trajectory over time, and identifying , individually and based on the at least one displacement characteristic, one or more vessels present within the vasculature.

[0006] In at least one embodiment in accordance with any previous / other embodiment described herein, filtering the ultrasound data comprises applying a long ensemble singular value decomposition (LE-SVD) clutter filter to the ultrasound data to separate the background signals from the reflected ultrasound signals.

[0007] In at least one embodiment in accordance with any previous / other embodiment described herein, the LE-SVD clutter filter has an ensemble size greater than 5,000 frames.

[0008] In at least one embodiment in accordance with any previous / other embodiment described herein, at least some of the one or more vessels have a diameter below about 100 pm.

[0009] In at least one embodiment in accordance with any previous / other embodiment described herein, the at least some of the one or more vessels are capillaries having a diameter below about 5 pm.

[0010] In at least one embodiment in accordance with any previous / other embodiment described herein, tracking the entire trajectory of the one or more ultrasound contrast agents through the vasculature comprises determining a position of each of the one or more ultrasound contrast agents in each timeframe associated with the reflected ultrasound signals, and generating a plurality of tracks indicative of the position of the one or more ultrasound contrast agents as determined.

[0011] In at least one embodiment in accordance with any previous / other embodiment described herein, the at least one displacement characteristic is a change in a velocity behavior of each of the one or more ultrasound contrast agents, the change in the velocitybehavior determined by filtering the plurality of tracks, spline interpolating the plurality of tracks, and derivating the plurality of tracks with respect to time to estimate a velocity of each of the one or more ultrasound contrast agents.

[0012] In at least one embodiment in accordance with any previous / other embodiment described herein, identifying the one or more vessels individually and based on the at least one displacement characteristic comprises determining that a respective one of the one or more ultrasound contrast agents exhibits a U-shaped velocity profile when traversing a respective one of the one or more vessels.

[0013] In at least one embodiment in accordance with any previous / other embodiment described herein, the method further comprises using the at least one displacement characteristic of each of the one or more ultrasound contrast agents to train at least one hidden Markov model to identify the one or more vessels.

[0014] In at least one embodiment in accordance with any previous / other embodiment described herein, the method further comprises detecting an occurrence of at least one vessel stall in the vasculature based on the at least one displacement characteristic of each of the one or more ultrasound contrast agents.

[0015] In at least one embodiment in accordance with any previous / other embodiment described herein, the method further comprises generating a reconstruction of an entire vascular network of the body based on the entire trajectory of the one or more ultrasound contrast agents through the vasculature, and outputting the reconstruction.

[0016] In at least one embodiment in accordance with any previous / other embodiment described herein, the method further comprises generating, based on the identifying, at least one two-dimensional (2D) super-resolution ultrasound localization microscopy (ULM) image of the vasculature and / or generating at least one pseudo-three-dimensional (3D) super-resolution ULM image of the vasculature, the at least one pseudo-3D superresolution ULM image providing elevational information about the vasculature, and outputting the at least one 2D super-resolution ULM image and / orthe at least one pseudo- SD super-resolution ULM image.

[0017] In at least one embodiment in accordance with any previous / other embodiment described herein, the method further comprises generating, based on the identifying, atleast one Contrast-enhanced ultrasound (CEUS) image of the vasculature, and outputting the at least one CEUS image.

[0018] In accordance with another aspect, there is provided a system for vasculature characterization using ultrasound. The system comprises a processing unit and a non- transitory memory communicatively coupled to the processing unit and comprising computer-readable program instructions executable by the processing unit for continuously acquiring ultrasound data from a body comprising a vasculature with one or more ultrasound contrast agents flowing therethrough, filtering the ultrasound data to separate background signals from reflected ultrasound signals generated by the one or more ultrasound contrast agents, tracking, based on the reflected ultrasound signals, an entire trajectory of each of the one or more ultrasound contrast agents through the vasculature, determining, based on the tracking, at least one displacement characteristic of each of the one or more ultrasound contrast agents throughout the entire trajectory over time, and identifying, individually and based on the at least one displacement characteristic, one or more vessels present within the vasculature.

[0019] In at least one embodiment in accordance with any previous / other embodiment described herein, the instructions are executable by the processing unit for filtering the ultrasound data comprising applying a long ensemble singular value decomposition (LE- SVD) clutter filter to the ultrasound data to separate the background signals from the reflected ultrasound signals.

[0020] In at least one embodiment in accordance with any previous / other embodiment described herein, the instructions are executable by the processing unit for tracking the entire trajectory of the one or more ultrasound contrast agents through the vasculature comprising determining a position of each of the one or more ultrasound contrast agents in each timeframe associated with the reflected ultrasound signals, and generating a plurality of tracks indicative of the position of the one or more ultrasound contrast agents as determined.

[0021] In at least one embodiment in accordance with any previous / other embodiment described herein, the instructions are executable by the processing unit for determining the at least one displacement characteristic comprising a change in a velocity behavior of the one or more ultrasound contrast agents, the change in the velocity behaviordetermined by filtering the plurality of tracks, spline interpolating the plurality of tracks, and derivating the plurality of tracks with respect to time to estimate a velocity of each of the one or more ultrasound contrast agents.

[0022] In at least one embodiment in accordance with any previous / other embodiment described herein, the instructions are executable by the processing unit for identifying the one or more vessels individually and based on the at least one displacement characteristic comprising determining that a respective one of the one or more ultrasound contrast agents exhibits a U-shaped velocity profile when traversing a respective one of the one or more vessels.

[0023] In at least one embodiment in accordance with any previous / other embodiment described herein, the instructions are executable by the processing unit for using the at least one displacement characteristic of each of the one or more ultrasound contrast agents to train at least one hidden Markov model to identify the one or more vessels.

[0024] In at least one embodiment in accordance with any previous / other embodiment described herein, the instructions are executable by the processing unit for at least one of generating a reconstruction of an entire vascular network of the body based on the entire trajectory of the one or more ultrasound contrast agents through the vasculature, and outputting the reconstruction, generating at least one two-dimensional (2D) superresolution ultrasound localization microscopy (ULM) image of the vasculature and / or generating at least one pseudo-three-dimensional (3D) super-resolution ULM image of the vasculature, the at least one pseudo-3D super-resolution ULM image providing elevational information about the vasculature, and outputting the at least one 2D super-resolution ULM image and / or the at least one pseudo-3D super-resolution ULM image, and generating at least one Contrast-enhanced ultrasound (CEUS) image of the vasculature, and outputting the at least one CEUS image.

[0025] Many further features and combinations thereof concerning embodiments described herein will appear to those skilled in the art following a reading of the instant disclosure.DESCRIPTION OF THE FIGURES

[0026] In the figures,

[0027] Fig. 1 is a block diagram of an example of a system for vasculature characterization using ultrasound, in accordance with one embodiment;

[0028] Fig. 2 is a block diagram of the image acquisition unit of Fig. 1 , in accordance with one embodiment;

[0029] Fig. 3A is a schematic diagram of a synthetic mouse brain microvascular network, in accordance with one embodiment;

[0030] Fig. 3B is a schematic diagram of a skeletonized vascular network with a single full vascular path identified, in accordance with one embodiment;

[0031] Fig. 3C is a schematic diagram of a zoomed portion of the vascular path of Fig. 3B;

[0032] Fig. 3D is a schematic diagram of a fully populated dataset of microbubbles moving through the mouse cerebral microvasculature of Fig. 3A, in accordance with one embodiment;

[0033] Fig. 3E is a graph of a velocity profile of a single microbubble, in accordance with one embodiment;

[0034] Fig. 4A is a graph of a representative power spectrum and the respective mean power spectrum as a function of sorted singular values, obtained for short ensemble singular value decomposition (SE-SVD) with an ensemble size of 800 frames and a frame rate of 450 Hz, SE-SVD with an ensemble size of 600 frames and a frame rate of 900 Hz, and long ensemble SVD (LE-SVD) with an ensemble size of 6,000 frames and a frame rate of 900 Hz, in accordance with one embodiment;

[0035] Fig. 4B is a graph of tracking results for a singular vascular path over the SVD clutter filtering conditions of Fig. 4A, in accordance with one embodiment;

[0036] Fig. 4C is a graph of the velocity profiles for the tracking results of Fig. 4B, in accordance with one embodiment;

[0037] Fig. 4D is a graph of an ideal U-shaped capillary velocity profile in situ, in accordance with one embodiment;

[0038] Figs. 5A and 5B illustrate diffusivity measurements obtained for selected capillary tracks in the brain cortex using the system of Fig. 1 , in accordance with one embodiment;

[0039] Figs. 6A, 6B, 6C, 6D, and 6E illustrate results obtained for a comprehensive mapping of the whole brain using the system of Fig. 1 in vivo, in accordance with one embodiment;

[0040] Figs. 7A, 7B, 7C, and 7D illustrate results obtained using the system of Fig. 1 as a function of Lipopolysaccharide (LPS) injection time, in accordance with one embodiment;

[0041] Figs. 8A to 8K illustrate results obtained using the system of Fig. 1 for quantifying capillary dwell time and stalling throughout the whole brain during neuroinflammation, in accordance with one embodiment;

[0042] Figs. 9A to 9H illustrate results obtained using the system of Fig. 1 for building individual vascular graph networks of veins and arteries, in accordance with one embodiment;

[0043] Fig. 10 is a flowchart of an example method for vasculature characterization using ultrasound, in accordance with one embodiment; and

[0044] Fig. 1 1 is a block diagram of an example computing device, in accordance with one embodiment.

[0045] It will be noted that throughout the appended drawings that like features are identified by like reference numerals.DETAILED DESCRIPTION

[0046] The field of ultrasound has undergone an unprecedented transformation with the advent of programmable ultrasound. In particular, using ultrasound localization microscopy (ULM), it has become possible to perform in vivo imaging of organs (e.g., the brain) of small animals at resolutions of the order of a few microns deep in tissue. ULM involves the injection of microbubbles into the vasculature of an image organ, the microbubbles remaining confined to blood vessels. As used herein, the term “microbubbles” refers to micron-sized contrast agents comprising aqueous suspensions in which bubbles of gas are bounded at the gas / liquid interface by a thin envelope involving an amphiphilic material (or surfactant). Such microbubbles typically have a size rangingbetween about 1 pm and about 5 pm, which is much smaller than the wavelength of ultrasound, which typically ranges between about 100 pm and about 800 pm. By leveraging the circulation of the free-flowing microbubbles within the organ’s vasculature, ULM enables the reconstruction of super-resolution images detailing micron-level vessel structures and blood flow velocities. As used herein, the terms “super-resolution” or “high resolution” refer to a resolution that is enhanced relative to the resolution attainable by a given imaging system. For example, super-resolution ultrasound images may refer to images with resolution that is finer than the diffraction limit.

[0047] Despite recent developments into extracting functional information beyond structural imaging or flow velocity using ULM, there remains a need to achieve single capillary resolutions (e.g., smaller than about 100 pm), to reduce the dependency of ULM on multiple microbubble track aggregates for vessel confirmation, and to simplify microbubble detection in capillaries. As used herein, the term “capillaries” refers to the microvascular unit (e.g., having a diameter below about 5 pm) which is at the interface between neurons and the circulatory system and which connects the input arterial networks to the output venous networks, serving as an important biomarker for neural health. Although reference is made herein to the systems and methods described herein being used for microbubble detection and characterization of capillaries, it should be understood that any vessel below the resolution of standard ULM may apply.

[0048] There are described herein methods and systems that may be used for non- invasive imaging and measuring of capillary function throughout an entire organ (e.g., the brain) at the single-capillary level. An ultrasound contrast agent is first administered to the vasculature of a subject. In one embodiment, the ultrasound contrast agent comprises microbubbles. It should however be understood that, while reference is made herein to the administration and the detection of microbubbles, any other suitable ultrasound contrast agent configured to be tracked (based on continuously acquired ultrasound data, as described further below) over a timeframe of seconds, including, but not limited to, nanodroplets and phase-changing droplets, may apply. In addition, the microbubbles may have linear or non-linear properties associated therewith.

[0049] In one embodiment, high frame rate ultrasound imaging is then used to sample microbubble flow. Although reference is made herein to the use of high frame rates, it should be understood that low frame rates may also be used, depending on theapplication. The data may be processed via super-resolution localization, tracking, and clutter filtering algorithms (e.g., long-ensemble singular value decomposition or LE-SVD). The result of this processing is a set of extended tracks of microbubble centers. As used herein, the term “microbubble track” refers to any microbubble (or other ultrasound contrast agent) seen and tracked over time and which can be used to generate a trajectory. The duration of the tracks is selected to enable, for each microbubble individually, tracking of the entire microbubble behavior (including that of slow moving microbubbles which are not erased as a result of the filtering process) through the vascular path through an extended timeframe for use in functional imaging. For example, the complete movement or trajectory of each microbubble, including the microbubble’s entry into a given arteriole, flow through a given capillary, and exit at a given venule, may be tracked. As used herein, the term “entire trajectory”, when used in relation to an ultrasound contrast agent (e.g., a microbubble), therefore refers to the entire movement of the contrast agent through the subject’s vasculature, from the contrast agent’s entry into the vasculature to its exit from the vasculature. It should be understood that the duration may vary depending on the application. In some embodiments, the tracks extend for several seconds (e.g., up to about four (4) seconds).

[0050] The change in behavior (e.g., velocity behavior) of the microbubbles over time is then determined from the tracks and used to identify one or more single vessels (e.g., single capillaries individually) present within the subject’s vasculature and thus characterize the vasculature. As used herein, the term “velocity behavior” refers to the movement of an ultrasound contrast agent (e.g., microbubble) or its position in space and time, allowing the assessment of local speed variations. The change in microbubble behavior (e.g., the change in velocity) may indeed be used to detect transitions between vasculature structures, and to accordingly identify these structures. As used herein, the term “identify” (when used in relation to vascular segment(s), vessel(s) or other structure(s) of a subject’s vasculature) refers to the fact that individual discrimination between the vascular segment(s) or vessels) (e.g., between arteries, veins, capillaries, etc.) can be achieved and that it is possible to establish what each detected vessel is (e.g., to recognize and name the vascular segment(s) or vessel(s) individually). The single vessels may be individually identified by training statistical hidden Markov models (HMMs) or using any other suitable technique. In particular, it is proposed herein to use temporal reporters referred to herein as Single capillary reporters (SCaRe) to identify anddiscriminate singular capillary tracks. As used herein, the term “capillary track” refers to a specific track that fits the a priori characteristics of a capillary based on the velocity behavior. In other words, and as will be described further below, temporal up-sampling is used for input into spatiotemporal singular value decomposition clutter filters to recover velocity profiles that correspond to flow at the capillary level.

[0051] The identified single capillary tracks may then be accumulated to generate outputs, such as maps of capillary function and microperfusion or other suitable outputs. In one embodiment, using the systems and methods described herein, it may be possible to obtain ULM images that achieve sub-diffraction-limited resolution (i.e. to image vessels below the diffraction limit), enabling the precise localization of individual capillaries and assessment of their structure and function. In particular, ULM images may be enhanced using the quantitative data (e.g., metrics) determined based on tracking of the microbubbles.

[0052] It should however be understood that, while reference is made herein to ULM images being generated and output using the systems and methods described herein, any other suitable image(s) may be generated. For instance, the systems and methods described herein may be used to generate one or more Contrast-enhanced ultrasound (CEUS) images of the subject’s vasculature, particularly for identifying vessels largerthan capillaries (e.g., arteries). Thus, CEUS images may also be enhanced using the quantitative data determined based on tracking of the microbubbles. It should therefore be understood that, while reference is made to the systems and methods described herein being applicable to small vessels (e.g., capillaries), the systems and methods described herein may also be applicable to large vessels (e.g., arteries, veins) such that both the subject’s microvasculature and the subject’s microvasculature can be characterized by tracking microbubbles (or other suitable ultrasound contrast agents) in the manner described herein.

[0053] It should also be understood that, in some embodiments, the systems and methods described herein may be used to track the movement of individual microbubbles (or any othersuitable contrast agent(s)) throughout multiple layers of a subject’s vasculature (e.g., through capillaries), identify vessel(s) based on the tracking, and assess metrics (e.g., capillary heterogeneous transit (CHT) time) based on the tracking and the vessel identification, without any image (e.g., ULM or CEUS) being formed.

[0054] Referring to Fig. 1 , there is illustrated a system 100 for vasculature characterization using ultrasound, in accordance with one embodiment. In the illustrated embodiment, the system 100 is composed of an ultrasound imaging device 102 and at least one probe 104. The ultrasound imaging device 102 is configured to operate at any suitable frequency range for imaging microbubble movement through the vascular network (i.e. through arterioles, to capillaries, and to venules) of a subject who has been administered an ultrasound contrast agent (e.g., microbubbles). In some embodiments, the ultrasound imaging device 102 is composed of two separate units, namely a wave generation unit 110 and an image acquisition unit 116. Similarly, the probe 104 may be composed of two separate probes, namely an emission probe 112 and a detection probe 114. When the system 100 is for CEUS imaging, a small probe 104 (e.g. having five (5) elements or the like) may be used. As such, the system 100 may be composed of two ultrasound devices 106, 108, one ultrasound device 106 (comprising the wave generation unit 110 and the emission probe 112) for wave generation and emission, and one ultrasound device 108 (comprising the detection probe 114 and the image acquisition unit 116) forwave detection and image reconstruction. Alternatively, all system components 110, 1 12, 1 14, 116 may be provided in a single housing as one device.

[0055] The wave generation unit 1 10 is a structure configured to generate at least one ultrasound wave or pulse. For this purpose, the wave generation unit 1 10 may comprise one or more ultrasound sources arranged in any suitable manner and configured to generate the ultrasound wave(s) (or pulses(s)) with sufficient power to travel in a region of interest of a target body 118 of the subject. The target body 1 18 is a biological body of a human or a non-human animal, such as an organ (e.g., the brain) that includes microvasculature (e.g., capillaries) with administered microbubble contrast agent(s). The wave generation unit 110 may comprise one or more ultrasound sources configured to generate unfocused or focused ultrasound waves, diverged waves, spherical waves, cylindrical waves, plane waves, or combinations thereof. The one or more ultrasound sources may also be configured to generate spatially or temporally encoded pulses. Any suitable ultrasound source may apply. In one embodiment, crystals, such as piezoelectric crystals, that vibrate in response to an electric current applied to the crystal may be used as the ultrasound source.

[0056] The emission probe 112 is configured for transmitting the ultrasound wave(s) or pulse(s) generated by the wave generation unit 1 10 towards the target body 118. Inresponse to the ultrasound wave(s) (or pulse(s)) being applied to the target body 1 18, one or more reflected ultrasound signals are generated by the presence of the ultrasound contrast agent (e.g., microbubbles) flowing through the vascular network of the target body 118. In one embodiment, the emission probe 1 12 is constructed using a frequency that matches the resonant ultrasound frequency of the ultrasound contrast agent in order to enable imaging thereof, the frequency being selected for maximum transmission, maximum reflective performance, and lowest noise. For example, a frequency in a range from about 20 KHz to about 50 MHz may be used.

[0057] The detection probe 1 14 is specifically configured to continuously detect and acquire ultrasound data comprising the reflected ultrasound signals generated from the ultrasound contrast agent (e.g., microbubbles). As used herein, the terms “continuous” and “continuously”, when used in relation to data acquisition, implies that the acquisition is “gap-less”, i.e. with no temporal gap between acquisitions. In one embodiment, the detection probe 1 14 comprises a high-frequency transducer configured to capture the reflected ultrasound signals at-depth throughout the target body 1 18, especially signals from microbubbles flowing through capillaries in the target body 118. The detection probe 114 may include, but is not limited to, a high frequency two-dimensional (2D) probe (e.g., as a linear 15 MHz probe), a high frequency three-dimensional (3D) probe (e.g., an 8 / 3 MHz matrix probe), or any other suitable type of 2D and 3D probe. The detection probe 114 may be configured to implement a detection sequence for the imaging task at hand including, but not limited to, focused-beam line-by-line imaging, wide-beam multi-line acquisition imaging, compounding plane wave imaging, compounding diverging wave imaging, synthetic aperture imaging, and the like. The detection probe 114 may be configured to acquire the ultrasound data in any suitable format including, but not limited to, ultrasound radio frequency (RF) data, ultrasound in-phase quadrature (IQ) data, ultrasound envelope data, and the like. In addition, the ultrasound data may contain a temporal dimension (i.e. a dimension in the temporal direction along which ultrasound frames are collected), and one or more spatial dimensions including, but not limited to, a lateral dimension, an axial dimension, an elevational dimension, and combinations thereof.

[0058] The image acquisition unit 116 is configured to receive the ultrasound data detected by the detection probe 114 and to process (using any suitable processing circuit,not shown) the ultrasound data (e.g., for image reconstruction and ULM), as will be described further below.

[0059] Referring now to Fig. 2 in addition to Fig. 1 , the image acquisition unit 116 illustratively comprises an input unit 202, a filtering unit 204, a localization and tracking unit 206, a segmentation unit 208, an image reconstruction unit 210, and an output unit 212.

[0060] The input unit 202 is configured to continuously acquire the ultrasound data, comprising the reflected ultrasound signals generated from the ultrasound contrast agent (e.g., microbubbles). In some embodiments, the input unit 202 may be configured to obtain the ultrasound data directly from the detection probe 1 14. In other embodiments, the ultrasound data may be stored in memory (or any other suitable storage device) following detection thereof by the detection probe 114 and the input unit 202 may therefore be configured to obtain the ultrasound data by retrieving the latter from memory. In particular, ultrasound stacks may be acquired in a continuous manner where the ultrasound data is off-loaded from memory at a sustainable rate for acquiring the next ultrasound stack. The input unit 202 is then configured to provide the ultrasound data to the filtering unit 204 for processing of the ultrasound data.

[0061] The filtering unit 204 is configured to filter the ultrasound data in order to remove scattering or clutter therefrom. Such scattering or clutter may be the result of tissue motions induced by the subject’s cardiovascular system (e.g., heartbeat and pulsatile motion from arteries) and / or by the subject’s respiratory system. The amplitude of tissue motions may be significantly larger than the size of the capillary to be resolved, thereby introducing blurring of the resulting images and leading to inaccurate measurements. In addition, noise signals may be falsely marked as reflected ultrasound signals generated from the ultrasound contrast agent (e.g., microbubble signals), resulting in inaccurate capillary delineation. The filtering unit 204 is therefore configured to use a clutter filtering technique to filter the ultrasound data to isolate in the ultrasound data the microbubble signals from background signals, such as tissue signals.

[0062] In one embodiment, the filtering unit 204 is specifically configured to isolate the microbubble signals while maintaining both the high velocity information and the low velocity information (i.e. the entire velocity range) associated with the microbubble signals.As used herein, the term “low velocity” refers to a velocity state where the ultrasound contrast agent(s) (e.g., the microbubble(s)) move at a velocity of less than 4 mm / s, and the term “high velocity” refers to a velocity state where the ultrasound contrast agent(s) move at a velocity that is three (3) times the standard deviation of the mean velocity during the low velocity state. The low and high velocity can be set and adjusted to any suitable value, depending on the application.

[0063] The recovery of slow moving microbubbles can prove particularly useful for challenging environments (e.g., transcranial). For this purpose, the filtering unit 204 may be configured to use singular value decomposition (SVD)-based filtering, and more particularly LE-SVD. As understood by those skilled in the art, SVD clutter filtering decomposes a spatiotemporal composition of data into eigenspaces (e.g., skull and brain tissue) and removal of singular values leaves decorrelated signals remaining (i.e. blood and noise). A singular value cutoff (also referred to as an “eigenvalue threshold”) is used to separate the background signals (which are typically projected to low-order singular values) from the microbubble signals (which are typically projected to intermed iate-to- high-order singular values). Using existing techniques, the cutoff between blood signal and stationary tissue is often overlapped and becomes more difficult to distinguish in situations with large motion artifacts, due to breathing or pulsation. Thus, in practice, the eigenvalue threshold is usually set sufficiently high as to remove the majority of stationary tissue. This however results in the removal of microbubbles moving at very low velocities, and thus the space surrounding larger penetrating arterials and veins where capillaries should appear dark in reconstructed images. In order to improve the discrimination between stationary tissue signal (e.g., brain and skull) and microbubble signal (e.g., slow moving microbubbles traversing capillary networks) and improve the recovery of tracks corresponding to capillaries, it is proposed herein to use temporal up-sampling for input into spatiotemporal clutter filters (e.g., SVD filters). In particular, it is proposed herein to increase the temporal sampling of SVD to implement LE-SVD with an ensemble size of thousands of frames. In one embodiment, ensemble sizes greater than about 5,000 frames at sufficiently high frame rates (e.g., greater than or equal to 900 Hz) to sample high velocity microbubble movement may be used. For example, LE-SVD with an ensemble size of about 6,000 frames may be used.

[0064] Although reference is made herein to the filtering unit 204 applying LE-SVD, it should be understood that any other suitable technique may apply and may be used incombination with LE-SVD (e.g., to enhance the filtering process). For example, in some embodiments, non-linear imaging may be used. In particular, non-linear imaging may be used to receive at twice the transmit frequency. This may enable higher resolution images of microbubbles (i.e. smaller point spread functions or PSFs).

[0065] The filtered data (also referred to herein as “microbubble signal data”) generated by the filtering unit 204 is then provided to the localization and tracking unit 206, which is configured to localize the ultrasound contrast agent (e.g., individual microbubbles) and track its movement through time using any suitable technique. Localization of the microbubbles may be performed by identifying, in each timeframe of the microbubble signal data, positions at which microbubbles are located. The microbubble locations may be defined in an (x, y, z) coordinate system (not shown), where the x axis represents the lateral direction, the y axis represents the elevational direction, and the z axis represents the axial direction. For instance, the center of each microbubble may be located (e.g., the center’s location in the lateral, elevational, and axial directions may be determined) such that the movement of the microbubble may be tracked through time. In some embodiments, a spatiotemporal tracking algorithm incorporating vesselness filtering, centerline thinning, and radial symmetry may be used to track (i.e. follow the movement of) microbubbles in space and time (e.g., over extended periods of time compared to conventional ULM). In particular, IQ data may be filtered via a Hessian-based vesselness filter that operates over space and time, specifically tuned to the expected PSF of a microbubble. Subpixel localization may be performed with a priori knowledge of the centers of the microbubble (given by the centerline) and updated via radial symmetry with a posteriori information given by the IQ data. A set of tracks indicative of the microbubble positions may be generated by the localization and tracking unit 206. The tracks may be Kalman filtered, spline interpolated, and derivated to produce velocity estimates.

[0066] The procedure implemented by the localization and tracking unit 206 will be described in further detail below, in accordance with one embodiment. This procedure may facilitate direct access to the spatiotemporal signal and velocity of the track, which was utilized for HMM categorization. Modifications were made to the vesselness filter to account for the temporal component in constructing vessel profiles, and the spatiotemporal tracker was adjusted to include track pairing. Backscattering amplitude was then recorded for each track as the absolute value of the IQ and the mean value of all the tracks per pixel were subsequently used for display on ULM and SCaRe maps.

[0067] Initially, the starts and ends of each track were paired if they were equivalent points in space and time. Subsequently, a second track pairing was performed to chronologically pair tracks with a distance (in x and z) less than 1 A with a temporal gap of less than 100 ms. Additionally, a Kalman filter was implemented immediately after radial symmetry localization. The Kalman filter was initialized with the first coordinate of the track and updated using the following equations for prediction:

[0068] where F is the state transition matrix, P is the identity matrix, and Q is the process noise covariance o. a was set empirically. These matrices are initialized as:

[0069] And the update steps are:

[0070] Where v is the estimated 2D coordinate, u is the actual 2D measurement (x and z coordinates), g is the desired grid spacing for normalizing the discrete values output by the centerline thinning algorithm, H is the measurement matrix, R, is the measurement noise e covariance, S is the residual covariance, K is the Kalman gain, and y is a scaling factor based on grid spacing. H and R are initialized as:

[0071] After the initial pass-through prediction and update, a Rauch-Tung-Striebel (RTS) smoothing filter was applied in reverse according to the following equations:

[0072] where C is the smoother gain, x and P are the updated state and covariance after smoothing. Capillary HMM training was conducted at this stage in the pipeline. Subsequently, tracks were spline interpolated and derivative with respect to time was taken for velocity display on ULM maps. A direction filter was applied to tracks to differentiate the colormap for upward movement (in one color, e.g. red) and downward movement (in another color, e.g. blue) based on the sign of the velocity in the z direction. Finally, these tracks were interpolated and accumulated on a super-resolved grid of A / 32, and the velocities were displayed over the track density map.

[0073] The segmentation unit 208 is configured to receive the set of tracks from the localization and tracking unit 206 and to process the tracks (a process referred to herein as SCaRe processing) in order to recover tracks that correspond to capillaries. In one embodiment, in order to classify capillary trajectories, the segmentation unit 208 is configured to measure the symmetry in each track (indicating the entrance and exit of a bubble through a capillary). The segmentation unit 208 is further configured to fit two exponentials to the beginning and ends of the trajectories. The R2value may be used to assess the likelihood that the track corresponds to a capillary. The segmentation unit 208 may also be configured to split the tracks into full capillary tracks, starts, and ends, where the starts are characterized by an exponential decrease in the velocity and the opposite for the ends. As understood by those skilled in the art, different travel paths may be followed by a microbubble through the subject’s vasculature. Microbubbles that are in penetrating veins and arteries, regardless of in and out of imaging plane travel, are expected to have relatively constant velocity throughout their travel. Microbubbles that enter capillaries are expected to have to slow down and speed up when they exit these meshes (i.e. exhibit an oscillatory behaviour). Additionally, it may be assumed that capillaries will not have large travel distances since their function is to provide nutrients to the surrounding perfused brain tissue. The segmentation unit 208 may therefore beconfigured to recover full tracks which contain velocity profiles that decrease then increase in a particular pattern (referred to herein as a “U-shape”), and to recover tracks that either start or end fast, indicating the entrances and exits of capillary tracks. Such a segmentation may be used to detect single capillaries and classify capillary trajectories. The capillary trajectories can then be converted into individual vascular graph networks of veins and arteries, as will be described further below.

[0074] While reference is made herein to identifying vessel(s) present in the subject’s vasculature based on the velocity behavior of the ultrasound contrast agent(s), it should be understood that any other suitable displacement (or travel) characteristics) of the ultrasound contrast agent(s) may be considered. For instance, the microbubbles’ acceleration may be determined from their velocity behavior (i.e. by computing a derivative of the velocity, which is itself determined by computing a derivative of the positions of the microbubbles) and used to identify the vessel(s). In addition, the mean squared displacement associated with the microbubbles may also be determined from the velocity behavior and used to identify the vessel(s). Diffusion coefficients may also be used (as will be discussed further below with reference to Figs. 5A and 5B). As understood by those skilled in the art, diffusion describes the manner in which particles move through biological tissues. In super-resolution optical imaging, particle tracking and estimation of the diffusion coefficient can reveal whether a particle exhibit directed flow, Brownian motion, or a combination of both. Such motion profiles facilitate understanding of the underlying biological processes (e.g., active transport mechanisms, cellular barriers, or pathological changes in tissue microenvironments) thereby providing insight into tissue function, health, and disease. Applying optical particle tracking techniques (such as the techniques described herein) for measuring diffusion may offer quantitative measurements of the function of microbubble transport through a capillary network and thus, may be useful in investigating the underlying pathologies in neurodegenerative and cardiovascular diseases.

[0075] In one embodiment, the process (i.e. SCaRe processing) implemented by the segmentation unit 208 encompasses HMM training, state prediction, and capillary categorization. It should however be understood that other embodiments may apply. In this embodiment, all tracks initially undergo preprocessing to convert the microbubble displacement characteristics (e.g., velocity profiles) into observations comprising a number of states. The states may be determined in any suitable manner, based on thedisplacement characteristics. In one embodiment, the states are determined based on microbubble velocity and acceleration. For example, four (4) states may be considered, namely high velocity and low acceleration, high velocity and high acceleration, low velocity and low acceleration, and low velocity and high acceleration Subsequently, all observations are standardized to the same length to facilitate HMM training (e.g., using the Baum-Welch algorithm implemented in MATLAB’s hmmtrain function). The initial transition probabilities are represented by a 2 x 2 matrix, while the emission probabilities are structured as a [2 x 4] matrix. In a second stage, state predictions on tracks are generated (e.g., using the Viterbi algorithm (hmmviterbi)), estimating whether the profile corresponds to a high-velocity state or a low-velocity state indicative of capillary flow. Finally, in a third stage, the states are characterized based on the occurrence and location of low-velocity states. Specifically, identification of whether the capillary network exhibits a U-shaped profile with a minimum CHT time threshold (e.g., of 0.120 s) is performed. As used herein, the term “CHT time” refers to the duration spent by a microbubble in the low- velocity state between the inlet and outlet points of the subject’s vascular network.

[0076] It should be understood that, while reference is made herein to HMM training, any other suitable technique may be used. For example, artificial intelligence (Al) and / or machine learning (ML) techniques may be used. It should also be understood that, since the approach proposed herein depends on the estimated velocity profile, this approach remains agnostic, which may prove useful for a variety of applications, including, but not limited to, identifying glomeruli and assessing function in a variety of organs.

[0077] The data produced as an outcome of the processing performed by the segmentation unit 206 may then be provided to the image reconstruction unit 210 for generation of one or more images (e.g., ULM super-resolution image(s) or CEUS image(s)) of the subject’s vasculature including the single vessel(s) (e.g., single capillaries) identified by the segmentation unit 206. As previously noted, the systems and methods described herein may be used to track the movement of individual contrast agents throughout multiple layers of a subject’s vasculature, identify vessel(s), and assess metrics without any image (e.g., ULM or CEUS) being formed. It should therefore be understood that the image reconstruction unit 210 is an optional element of the image acquisition unit 116 and may not be used in all embodiments.

[0078] In one embodiment, based on the data received from the segmentation unit 208, the image reconstruction unit 210 may be configured to generate 2D images of the vasculature (i.e. images in the lateral-axial, or (x, z), plane), as well as pseudo-3D images (i.e. images in the lateral-elevational-axial, or (x, y, z), space) which additionally provide elevational information (i.e. data along the y axis representative of the elevational direction) about the vascular network being imaged. It should however be understood that the systems and methods described herein may be implemented in embodiments where an underlying image, which is already in 3D, is used. In other words, the systems and methods described herein may be used for 2D imaging and 3D imaging applications.

[0079] The image reconstruction unit 210 may implement any suitable technique to generate the images, which are then sent by the image reconstruction unit 210 to the output unit 212 to cause the images to be displayed or stored in memory (or other suitable storage device) for later use. For example, the output unit 212 may be configured to cause the images to be presented in a superimposed manner or side-by-side on a suitable output device (e.g., a computer screen). The output unit 212 may also be configured to cause the images to be stored in memory for subsequent analysis including, but not limited to, microvascular morphology measurements (e.g., vessel density and vessel tortuosity), hemodynamics measurements (e.g., blood flow speed and volume), and the like. Using the systems and methods described herein, it may also be possible to assess neuronal function via neurovascular coupling.

[0080] While reference is made herein to the output unit 212 receiving image(s) from the image reconstruction unit 210, it should be understood that, in some embodiments, the output unit 212 may be configured to alternatively or additionally receive any other suitable data (e.g., metrics determined by the segmentation unit 206) for output (e.g., rendering one display and / or storage in memory).

[0081] In one embodiment, for visualization purposes, capillary trajectories are processed by accumulating Gaussians centered at super localized track positions (e.g., by accumulating centered positions within the HMM Viterbi segmented regions of the super localized track positions). Both the intensity and size of the Gaussian functions are modulated by estimated capillary sections, ensuring that inlet and outlet points from arterioles / venules appear transparent and small while the capillaries have a larger spread. Additionally, intensity scaling by the frame rate enables the resulting map to highlightpixels where microbubbles remain stationary, indicating time spent in the capillary. In particular, in one embodiment, the SCaRe map shows capillary dwell time, an indirect measurement of transit-time, as the integral overtime given by the frame rate, highlighting pixels where microbubbles remain stationary, indicating time spent in the capillary given by:

[0082] Where t1 and t2 indicate the beginning and end of the ultrasound scan, N indicates the number of capillary tracks, and 6 is a Dirac delta function placed at the capillary track coordinates summated over time. Subsequently, SCaRe maps are superimposed onto greyscale ULM maps of all trajectories, effectively integrating functional and structural information for comprehensive visualization and analysis.

[0083] As will be described further below, in order to quantify capillary stalling, HMMs were used to classify possible tracks that enter or exit a capillary vessel. Each entrance track was paired across the scan time with an exit track only if they occur within an Euclidean distance of 10 pm of each other. The locations, the travel path, and the stalled times were recorded for each pair and a gaussian convolved map was overlaid onto ULM images to create stalling hotspot maps for ROI segmentation in histology images.

[0084] Computational models used to uncover the limitations in ULM, including the requisite acquisition length and the challenges in distinguishing slow-moving microbubbles, will now be described. In one embodiment, computational modeling of microbubble flow was performed within simulations of the mouse brain microvasculature with fully-connected closures of over millions of vessels. In particular, synthetic networks were used to emulate complete and balanced circulation, encompassing fully-connected capillary closures and topology from arteries to veins. This in situ model allowed to anticipate the behaviors of microbubbles within capillaries and proposed new strategies for overcoming existing limitations of exiting ULM techniques in capturing slow capillary flow.

[0085] By constructing a dataset of microbubbles flowing through the microvasculature, all feasible paths from inlet (i.e. arteries) to outlets (i.e. veins), traversing a single capillary, can be retrieved, as can be seen in Figs. 3A, 3B, and 3C. Fig. 3A shows a synthetic mouse brain microvascular network 300 with fully-connected capillary networks, blood flow, vessel diameter, and pulse pressure, in accordance with one embodiment. Fig. 3B shows a skeletonized vascular network 310 with a single full vascular path identified. Fig. 3C shows a zoomed image 320 of vascular path with vascular components color-coded. Although a simplified representation of the capillary mesh, the graph network provides insight into the overall movement and microbubble velocity throughout the capillary network, especially with access to metrics such as blood flow, vessel radius, pulse pressure, CHT, and capillary stalling. In particular, and as will be described further below, using the systems and methods described herein, the occurrence at least one vessel stall (e.g., a capillary stall) in the vasculature can be detected based on the displacement characteristics (e.g., velocity behavior) of the microbubbles.

[0086] As further illustrated in Figs. 3A, 3B, and 3C, sequential Monte Carlo simulations, with random sampling of specific paths and initial conditions, were built to simulate microbubbles for further ultrasound data simulation. Specifically, random sampling was performed from a simulated distribution of modeled Definity microbubbles (mu = 2 pm, std = 3 pm). For one hemisphere, this approach yielded 291 ,372 possible paths through capillaries. Thus, each microbubble was forward simulated (i.e. forward propagated from inlet to outlet) taking into account the size of the bubble, the flow velocity (calculated from instantaneous blood flow and vessel diameter), as well as the pulse wave velocity, simulated as a traveling wave of 350 cm / s at 500 beats per minute (BPM). Additional constraints were put so that a bubble cannot pass through a capillary smaller than the diameter. Afterwards, microbubble datasets were stitched together, synchronous with the cardiac cycle, to produce a fully populated dataset of microbubbles moving through the mouse cerebral microvasculature (see graph 330 Fig. 3D). The key insight was the discovery of distinct U-shaped velocity behaviors as a single microbubble traverses with low velocities at the capillary mesh (see velocity profile 340 of Fig. 3E). It should be understood that the U-shape exhibited by the velocity profile may be formed of a single microbubble trajectory (as illustrated in Fig. 3E for instance) or of multiple microbubble trajectories connected to one another. For example, the U-shape may be formed by connecting two (2) J-shaped microbubble trajectories.

[0087] To further simulate ultrasound data from the microbubble dataset, the microvasculature was aligned with an open source repository for mouse micro-computed tomography (or micro-CT). Then, a 3D portion of the microbubbles and skull micro CT were segmented to fit within the elevational and lateral footprint of a 16 MHz 128-element probe (spacing = 100 pm, element height = 1.5 mm). Twelve (12) scatterers per resolution cell (A 14) were simulated and arranged according to the intensity of the micro CT so that the scatterers placed at the skull were of higher concentration than the brain matter. These scatterers were superimposed with the centers of the microbubbles for input into a linear ultrasound simulator. A GPU-accelerated simulator based on the equations from the ultrasound simulator SIMUS was used to simulate IQ data for 3500 plane waves (7 angles between -11 ° and 11 °) at 7000 Hz PRF. Simulated results were stored so that the full IQ datasets were linearly superimposed RF from microbubbles and RF from skull clutter. The analysis of the simulated results allowed to discern limitations of SVD-based clutter filtering.

[0088] Due to temporal embeddings within the SVD algorithm, SVD clutter filtering negatively impacts the ability to recover capillary tracks in vivo. These temporal embeddings correspond to spatiotemporal ensembles of less than 1000 frames, which are formulated as a Casorati matrix and may embed slow moving microbubbles within the stationary tissue eigenspaces. A dual challenge is therefore posed by the time required for, and the speed at which, microbubbles traverse through a capillary network. This in turn hinders the recovery of U-shaped velocity profiles, where stationary microbubbles are eliminated. There is therefore a need for acquisition of continuous data to increase ensemble sizes to maximize the likelihood of capturing capillary tracks. Indeed, ULM sequences recorded non-continuously for less than a given timeframe (e.g., 4 seconds) may be insufficient to adequately sample capillary transit-times. In addition, the temporal sampling of SVD, which is crucial for distinguishing slowly moving microbubbles, is influenced by the frame rate and rank of the input matrix. Microbubbles traveling at slow speeds, approaching the stationary tissue regime, risk being attenuated in the SVD filtering process if temporal sampling is inadequate. Thus, in one embodiment and as described herein above, it is proposed herein to apply LE-SVD in vivo as a means to retain the full velocity spectrum necessary for recovering U-shaped microbubbles behaviors.

[0089] Fig. 4A illustrates a graph 400 of a representative power spectrum, and the respective mean power spectrum as a function of sorted singular values, obtained forthree different conditions in vivo, namely short ensemble SVD (SE-SVD) with an ensemble size of 800 frames and a frame rate of 450 Hz (referred to herein as “low frame rate SE- SVD”), SE-SVD with an ensemble size of 600 frames and a frame rate of 900 Hz (referred to herein as “high frame rate SE-SVD”), and LE-SVD with an ensemble size of 6,000 frames and a frame rate of 900 Hz. For each condition, the first 20 eigenvalues were eliminated to remove stationary tissue signal from the skull and brain tissue, leaving only the blood and microbubble signals. Microbubbles were injected via tail-vein catheterization and tracked using a spatiotemporal tracking methodology. The SVD frequency spectrum density shown in Fig. 4A indicates that there are distinct frequency regimes of stationary tissue and blood tissue after SVD, and that smoother transitions from tissue to blood regimes can be obtained using LE-SVD. As can be further seen from Fig. 4A, tissue signals remain concentrated at the first eigenvectors but LE-SVD condenses the spread of stationary tissue signals. This stretching of spectral density across large ensemble sizes in turn facilitates better separation between slow microbubble signals and stationary tissues.

[0090] Fig. 4B illustrates a graph 410 of representative tracking results with identical parameters (i.e. the same microbubble and vessel in the mouse brain) for a singular vascular path over the above-mentioned SVD clutter filtering conditions, i.e. SE-SVD with an ensemble size of 800 frames and a frame rate of 450 Hz, SE-SVD with an ensemble size of 600 frames and a frame rate of 900 Hz, and LE-SVD with an ensemble size of 6,000 frames and a frame rate of 900 Hz. From Fig. 4B it can be seen that, while low frame rates effectively recover slow-moving bubbles by increasing their displacement in time, they fail to capture the fast-moving transition into and out of the capillary, losing the entire trajectory in the process. Conversely, high frame rates capture high velocity but miss the microbubble at low velocities due to the insufficient sampling in the SVD, hindering full vessel tracking. In contrast, LE-SVD processed frames produce the entire estimated track, including entry, traverse, and exit from the capillary track.

[0091] Fig. 4C illustrates a graph 420 of the velocity profiles for the respective tracking results in Fig. 4B and Fig. 4D illustrates a graph 430 of the ideal U-shaped velocity profile in situ. It can be seen that distinctive U-shaped velocity profiles are exhibited by microbubbles traversing capillary vessels. SCaRe relies on the extraction of these U- shaped velocity profiles, depicting the high-velocity descent through arterioles, slow perfusion through capillaries, and high-velocity uptake into venules. While low frame rateSE-SVD shows a low velocity track prone to false pairing, high frame rate SE-SVD splits the velocity behavior into two tracks without indicating their association. Conversely, LE- SVD enables the recovery of the entire U-shaped velocity profile. The combination of LE- SVD and adaptive track pairing further increases the number of capillary tracks recovered. These results demonstrate the capability of LE-SVD in overcoming SVD limitations in tracking whole capillary networks in vivo.

[0092] Referring now to Figs. 5A and 5B, diffusivity measurements obtained for selected capillary tracks in the brain cortex will now be discussed, in accordance with one embodiment. As noted herein above, the techniques described herein may be used to measure diffusion and provide quantitative measurements of the function of microbubble transport through a capillary network. In the illustrated embodiment, ULM was performed using a 10 MHz linear array and capillary tracking was performed via SCaRe frameworks. Capillary tracks were sampled from a single six (6) second scan period. A sliding window of 25% of the total track length N was used to estimate the mean squared displacement (MSD) as a function of lag T in x and z directions, as follows:

[0093] Afterwards, the curve was fit with a line of which the slope can be retrieved, and the diffusion coefficient (D) can be calculated for each midpoint of the sliding window along the track, as follows: d lim — MSD(T)co dt

[0094] where d indicates the number of dimensions (two (2) in this case).

[0095] Figs. 5A and 5B show capillary tracks from one buffer (one (1) second of data) overlaid onto an ULM scan. In Fig. 5A, track color indicates the measured local diffusion coefficients derived from a sliding window mean squared displacement linear fit estimation. Fig. 5B is a zoomed panel displaying a U-shaped track 502 and J-shaped track 504 indicating the passage from arteriole to capillary to venule. The self-diffusivity results of Figs. 5A and 5B indicate that segments where microbubbles travel through capillary meshes have low diffusivity, indicating higher resistance through the confined capillaryvessels where oxygen exchange from red blood cells to surrounding cells takes place.This presents a quantitative biomarker for nutrient transport across capillary meshes.

[0096] Referring now to Figs. 6A, 6B, 6C, and 6D, results obtained for a comprehensive mapping of the whole-brain using SCaRe in vivo will now be discussed, in accordance with one embodiment. Continuous data acquisition of a bolus of Definity microbubbles (1 :10 dilution), enhanced clutter filtering using LE-SVD, and the implementation of HMMs to discern microbubble descent and ascent through capillaries were performed. Fig. 6A illustrates a velocity map 600 with extended tracks, accompanied by the corresponding SCaRe map beneath (Bregma = -1.5 mm). Notably, distinct single capillary tracks are discernible in both cortical and subcortical regions of the brain. Zoomed panels 610 in Fig. 6B isolate a single vascular path composed of a penetrating arteriole connected to penetrating venules. Remarkably, this map demonstrates the composite nature of three separate microbubble paths originating from the same arteriole. The ability to discern capillaries in this fashion demonstrates heterogeneity in the spatiotemporal patterns of capillaries in ULM, which was previously challenging using existing techniques.

[0097] Furthermore, Fig. 6C is a graph 620 that showcases the ability to segregate measured CHT by neuronal layers. This capability is particularly pertinent given existing studies indicating layer-specific differences in capillary function, notably in age-related white matter loss. Individual SCaRes can be analyzed to discern their travel path, velocity behaviors, and estimated CHT, as depicted in the graph 630 of Fig. 6D. Notably, capillaries identified using SCaRe exhibit high-velocity entry, followed by low-velocity capillary transit, and culminating in high-velocity uptake into penetrating venules. Notably, these capillaries were tracked over a minute, occurring seconds apart. The observation of differing capillary transits originating from the same arteriole and exiting through the same venule underscores the spatiotemporal complexity of microvascular dynamics. The ability to elucidate such patterns for single capillaries presents numerous avenues for advancing the understanding of brain health, such as how the brain ages as well as how neuronal function is impacted by capillary heterogeneity.

[0098] As described herein, SCaRe may be used as a correlate for capillary transit time. It should however be understood that SCaRe opens avenues for exploring other potential biomarkers, such as investigating heterogeneity within a single capillary mesh (as illustrated in Figs. 6C, 6D, and in the graph 640 of Fig. 6E) or understanding local capillaryfunction within specific brain regions or cortical layers. While the measured CTH in vivo (1.4 seconds) falls below the maximum CTH (4.0 seconds) observed in simulations, measurements appear consistent with those reported in the literature under steady-state conditions.

[0099] The validity of SCaRe biomarkers was validated by measuring the CHT within the same animal and imaging plane before and 1 and 2 hours after injection of Lipopolysaccaride (LPS). Fig. 7A illustrates images 700, 702, and 704 of representative results obtained using SCaRe ULM at timepoints of 0 hour, 1 hour, and 2 hours post LPS injection, respectively. By inducing global inflammation, lymphocytic activity was increased, increasing the number of capillary stalls in the brain, which is expected to increase CHT. After constructing ULM maps and quantifying SCaRe biomarkers, it is not immediately clear the influence of LPS on the distribution of measured CTH in the whole brain slice. The Kolmogorov-Smirnov one-tailed tests with Bonferonni corrections illustrated in the graph 710 of histogram plots of Fig. 7B revealed that in mouse 1 , the distribution of CTH at 2 hour was significantly different than at 1 hour and, although not significant, at baseline. In mouse 2, only the distribution of CTH at 1 hour was significantly different than baseline. The Kolmogorov-Smirnov test indicates that the distribution of CTH for both mice at baseline can be drawn from the same distribution. Fig. 7C illustrates a scatter plot 720 of baseline subtracted density counts for the distributions after LPS injection (time 1 hour and 2 hour). Here, a small trend is seen towards increase over the range of measured CTH in SCaRe. Nonetheless, the largest CTH recorded were consistently after LPS was injected. Lastly, CTH can be separated and analyzed locally via cortical-subcortical segmentation based on the Allen brain institute atlas (Common Coordinate Framework version 3). Fig. 7D is a graph 730 of histogram plots which shows that, for mouse 1 , a significant increase in the distribution tails in the cortex after LPS injection can be seen. This trend also holds true in the subcortex, but with t = 1 h almost reaching significance. Mouse 2 shows more significant distribution changes in the subcortex than the cortex with t = 1 h reflecting the changes seen globally (see Fig. 7B).

[0100] Referring now to Figs. 8A to 8K, the use of SCaRe biomarkers for quantifying capillary dwell time and stalling will now be illustrated. In particular, it was first investigated whether SCaRe biomarkers have the ability to detect sensitive changes in capillary function in normal physiology and under duress at the single capillary level throughout the whole brain. To induce neuroinflammation, an intraperitoneal (IP) lipopolysaccharide(LPS) challenge was used to increase neutrophil circulation and subsequent capillary stalling in n = 6 mice compared with n = 6 phosphate-buffered saline (PBS)-vehicle SHAM mice. Capillary dynamics were localized overthree time points, namely: baseline, 1 h-post, and 2h-post. The length of time that a microbubble spends in a capillary, co-registered with the Allen Brain Atlas (plate 83), was then quantified.

[0101] Fig. 8A shows a representative baseline-subtracted difference maps of mean capillary dwell time at timepoints 1 h and 2h after LPS-challenge. Fig. 8B shows the quantification of dwell time changes in maximum, mean, and standard deviation over the three previously mentioned timepoints. Fig. 8C shows the mean difference between timepoints and baseline probability density functions of recorded capillary dwell times for SHAM (see left side of Fig. 8C) and LPS (see right side of Fig. 8C) injection. Fig. 8D shows the heat-map of brain region analysis of capillary dwell times on the Allen Mouse Brain Atlas (Plate 83). Significant regions of interest (ROIs) for changes in maximum, mean, and standard deviation are denoted by an asterisk (* p < 0.05, paired t-test with false discovery rate correction). Fig. 8E shows significant brain areas in Fig. 8D with overlapping areas illustrated in stripes. Fig. 8F shows mean baseline-subtracted capillary dwell times for each ROI in Fig. 8E. Fig. 8G shows a visual representation of capillary stalling in 2D (see top of Fig. 8G) and 3D with respect to time (see bottom of Fig. 8G). Fig. 8H shows stalling occurrence at each timepoint over the whole scan time duration. Fig. 8I shows representative histology images of microglia stained with I ba 1 antibodies taken at stalling areas (see hotspot 802 in Fig. 8I) for SHAM and LPS mice. Microglia morphology using cluster analysis is overlaid on top of Fig. 8J and shows capillary stalling frequency over three timepoints (* p < 0.05, 2-way ANOVA multiple comparisons). Finally, Fig. 8K shows the percentage of microglia distributions within stalling ROIs (** p < 0.01 , ***p < 0.001 , Generalized Linear Mixed Models with Bonferroni Correction).

[0102] Fig. 8A demonstrates brain-wide changes in the average capillary dwell time at 1 h and 2h post LPS / SHAM injection from baseline recordings. An overall global change in LPS-challenge mice is seen in contrast to an overall decrease in capillary dwell time in PBS-injected mice. If global changes are quantified throughout the brain, it can be seen that neuroinflammation shows a monotonic increase in maximum, mean, and standard deviation metrics for LPS in contrast to a greater decrease in SHAM mice (see Fig. 8B). A two-way repeated measures ANOVA procedure indicates that changes in global dwell times over the three time points are not significantly different between groups (p = 0.094max; p = 0.083 mean; p = 0.063 std; Greenhouse-Geisser adjustment). Changes to capillary transit-time distributions within subjects are compared by calculating baseline- subtracted probability density functions for each animal. The average changes are illustrated in Fig. 8C where notable changes between the two groups are identified. In SHAM mice, a continuous shift in the distribution towards the left with increased fast transits and decreased slow transits is observed. However, LPS-challenge instead shifts the distribution rightwards, increasing the occurrence of larger transit-times while decreasing the probability of faster transit-times. This effect is greater at the 1 h timepoint than the 2h timepoint.

[0103] Systemic neuroinflammation is expected to change the neurovascular unit heterogeneously throughout the brain and changes in capillary transit times are expected to follow similar trends. If a regional analysis of the changes in each brain regions is performed, one can identify ROIs that experience significant changes. Using a two-way paired t-test with corrections for false discovery rates, the heatmap of mean dwell time changes in Fig. 8D denotes brain regions with significant changes with an asterisk. These areas are illustrated on the atlas ROI map in Fig. 8E with each color signifying either changes in maximum, mean, and / or standard deviation. Indeed, widespread, almost symmetrical changes, in both cortical and subcortical structures are seen. Moreover, it can be seen that hippocampal subregions such as CA1 and CA3 have significant neurovascular responses to neuroinflammation. Lastly, if baseline-subtracted mean dwell times for these regions are plot, one finds that brain regions during neuroinflammation are elevated in comparison to PBS-SHAM regions that show a large decrease over time.

[0104] Due to the large implications that capillary stalls can have on neuroinflammation, stroke, and neurodegeneration, it was further investigated whether SCaRe can be used to measure rare but important occurrences of stalling deep in the brain. The proposed HMM configuration described herein can identify microbubble high-low (arteriole to capillary) and low-high (capillary to venule) state changes which can be paired if the capillary segments occur within 10 microns of each other. These tracks can be temporally projected in 3D to visualize individual capillary stalls (see Fig. 8G) and their occurrence over the minute long scans can be depicted (see Fig. 8H). Furthermore, an association with neuroinflammation-activated microglia to capillary stalling by segmenting ROIs in I ba 1 immunohistology slices at stalling vessel sites in LPS and SHAM mice was elucidated and morphological cluster analysis of microglia states performed (see Fig. 8I). As expected, itwas observed that both SHAM and LPS animals experience natural capillary stalling. Quantifying the stalling frequency over the whole scan time at the three (3) different timepoints, one sees marked increase in LPS-challenge animals at 1 h with a decrease at 2h (see Fig. 8J). However, in SHAM animals, there are no large changes across all timepoints. A two-way ANOVA procedure with tests for multiple comparisons indicates that the interaction between LPS-challenge and timepoint is statistically significant (p = 0.019) with the most significant difference (p = 0.022) between the two groups at the 1 h timepoint. Importantly, analysis of the microglia morphology surrounding capillary stalls show changes in the proportions of activated microglia (see Fig. 8K). Statistical significance between LPS and SHAM microglia morphology can be observed with no changes in the proportion of active amoeboid and rod-like microglia. However, one sees statistically significant decreased proportions of ramified (surveillant, homeostatic) and increased proportions of hypertrophic microglia that play active roles in neuroinflammation.

[0105] Referring now to Figs. 9A to 9H, the use of SCaRe for performing segmentation will now be illustrated. As previously noted, SCaRe can be used for mapping capillary vessels from a single microbubble based on the transit from an arteriole to a venule. This unique information can then be leveraged to build vascular graph networks from a single ULM scan. Indeed, localized trajectories can be converted to individual vascular graph networks of veins and arteries. To evaluate SCaRe when used for segmentation purposes, IQ data (e.g., a 16 MHz linear array, 5 angles ± 5°, 5kHz) was acquired continuously (e.g., for a duration of five (5) minutes) with an ultrasound system (e.g., a Verasonics Vantage 256™ ultrasound system) with a 1 :5 dilution microbubble infusion (e.g., using Definity® microbubbles). As can be seen from Fig. 9A, track-and-localize reconstructs ULM up- down flow and SCaRe. Graph inference was performed via RoadRunner, a road-network GPS inference framework, to reconstruct individual artery-venous trees, based solely on trajectory information. Initial seeds were first sampled from skeletonized ULM maps (see Fig. 9B) then Waypath filtering (see Fig. 9C) - based on forward trajectories through linked nodes - reconstructed graph networks (see Fig. 9D).

[0106] ROI of Up-Down (see Fig. 9E), SCaRe vessel remapping (see Fig. 9F), and difference maps (see Fig. 9G) indicate large changes in cortical vessels but no changes in subcortical penetrating vessels. SCaRe-mapped complex flow (right column of Fig. 9F) in hippocampal arteries that have bidirectional flow is now mapped to the same artery.Graph inference (see Fig. 9D) demonstrates greater ability to connect vasculature than via skeletonization on binary segmented vessels from ULM (see Fig. 9B) with greater discrimination of vessels running in parallel (see arrows 902a and 902b in Fig. 9D). Two separated whole-brain graphs (red-artery, blue-vein) can then be overlaid with velocity information integrated into nodes and edges (see Fig. 9H). It can be seen that the segmentation method described herein not only enables the analysis of vasculature that are normally indistinguishable in ULM, but also recovers anatomically classified vessels in the whole brain.

[0107] Referring now to Fig. 10, there is illustrated an example method 1000 for vasculature characterization using ultrasound. The method 1000 may be performed by the ultrasound device 108 of Fig. 1 (e.g., by the image acquisition unit 116 of Fig. 1). Step 1002 comprises continuously acquiring ultrasound data from a body comprising a vasculature with one or more ultrasound contrast agents flowing therethrough. The ultrasound data may be obtained from the detection probe 114 in the manner described herein above. Step 1004 comprises filtering the ultrasound data to separate background signals from reflected ultrasound signals generated by the one or more ultrasound contrast agents. In one embodiment, LE-SVD filtering with an ensemble size greater than 5,000 frames is used at step 1004, as described above. Step 1006 comprises tracking, based on the reflected ultrasound signals, an entire trajectory of the one or more ultrasound contrast agents through the vasculature. In one embodiment, step 1006 entails determining a position of each of the one or more ultrasound contrast agents in each timeframe associated with the reflected ultrasound signals, and generating a plurality of tracks indicative of the position of the one or more ultrasound contrast agents as determined. Step 1008 comprises determining, based on the tracking, at least one displacement characteristic (e.g., a change in velocity behavior) of each of the one or more ultrasound contrast agents throughout the entire trajectory over time. In one embodiment, step 1008 entails filtering the plurality of tracks, spline interpolating the plurality of tracks, and derivating the plurality of tracks with respect to time to estimate a velocity of each of the one or more ultrasound contrast agents.

[0108] Step 1010 comprises identifying, individually and based on the at least one displacement characteristic, one or more vessels present within the vasculature. At least some of the one or more vessels have a diameter below about 100 pm. In one embodiment, these vessels are capillaries having a diameter below about 5 pm. In oneembodiment, identifying the one or more vessels based on the displacement characteristic(s) comprises determining that the one or more ultrasound contrast agents exhibit a U-shaped velocity profile when traversing the one or more vessels, as described herein above.

[0109] In some embodiments, the method 1000 further comprises using the displacement characteristic(s) of each of the one or more ultrasound contrast agents, as determined at step 1008, to train at least one hidden Markov model to identify the one or more vessels.

[0110] In some embodiments, the method 1000 further comprises generating at least one image of the vasculature including the one or more vessels as identified, and outputting the at least one image.

[0111] In some embodiments, generating the at least one image comprises generating a reconstruction of an entire vascular network of the body based on the entire trajectory of the one or more ultrasound contrast agents through the vasculature.

[0112] In some embodiments, generating the at least one image comprises generating at least one two-dimensional (2D) super-resolution ULM image of the vasculature and / or generating at least one pseudo-three-dimensional (3D) super-resolution ULM image of the vasculature, the at least one pseudo-3D super-resolution ULM image providing elevational information about the vasculature.

[0113] In some embodiments, generating the at least one image comprises generating at least one CEUS image of the vasculature.

[0114] In some embodiments, the occurrence of at least one vessel stall in the vasculature is detected based the displacement characteristic(s) of each of the one or more ultrasound contrast agents, as described herein above.

[0115] In some embodiments, parts or all of the method 1000 of Fig. 10 are performed by a computing device 1 100, as illustrated in Fig. 11. The computing device 1100 may thus be used to implement parts or all of the system 100 of Fig. 1 . The computing device 1100 comprises a processing unit 1102 and a memory 1 104 which has stored therein computerexecutable instructions 1 106. The processing unit 1102 may comprise any suitable devices configured to cause a series of steps to be performed such that instructions 1106, when executed by the computing device 1 100 or other programmable apparatus, maycause functions / acts / steps described herein to be executed. The processing unit 1102 may comprise, for example, any type of general-purpose microprocessor or microcontroller, a digital signal processing (DSP) processor, a CPU, an integrated circuit, a field programmable gate array (FPGA), a reconfigurable processor, other suitably programmed or programmable logic circuits, or any combination thereof.

[0116] The memory 1104 may comprise any suitable known or other machine-readable storage medium. The memory 1 104 may comprise non-transitory computer readable storage medium, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. The memory 1104 may include a suitable combination of any type of computer memory that is located either internally or externally to device, for example random-access memory (RAM), read-only memory (ROM), electro-optical memory, magneto-optical memory, erasable programmable read-only memory (EPROM), and electrically-erasable programmable read-only memory (EEPROM), Ferroelectric RAM (FRAM) or the like. Memory 1 104 may comprise any storage means (e.g., devices) suitable for retrievably storing machine-readable instructions 1106 executable by processing unit 1 102.

[0117] The method 1000 may be implemented, in part or entirely, in a high level procedural or object oriented programming or scripting language, or a combination thereof, to communicate with or assist in the operation of a computer system, for example the computing device 1100. Alternatively, the method 1000 may be implemented in assembly or machine language. The language may be a compiled or interpreted language.

[0118] Embodiments of the method 1000 may also be considered to be implemented by way of a non-transitory computer-readable storage medium having a computer program stored thereon. The computer program may comprise computer-readable instructions which cause a computer, or more specifically the processing unit 1102 of the computing device 1100, to operate in a specific and predefined manner to perform the functions described herein.

[0119] Computer-executable instructions may be in many forms, including program modules, executed by one or more computers or other devices. Generally, program modules include routines, programs, objects, components, data structures, etc., thatperform particular tasks or implement particular abstract data types. Typically, the functionality of the program modules may be combined or distributed as desired in various embodiments.

[0120] In some embodiments, the systems and methods described herein, by leveraging a combination of LE-SVD, spatiotemporal tracking, track pairing, and HMMs, may be used to successfully address the inherent limitations of ULM and accurately identified single capillaries amidst a forest of tracks. Unlike conventional techniques, the proposed modality may indeed be used to map single capillary flow and heterogeneity throughout a whole organ over time. In particular, the SCaRe methodology may achieve measurement of capillary heterogeneous transit time (CHT) and the mapping of microbubble trajectories based on their behavior as they enter and exit capillaries throughout an entire organ (e.g., the entire brain), a feat unachievable by existing optical, ultrasound, CT, or MRI-based neuroimaging modalities. In addition, LE-SVD may enable to recover smaller flow rates and trace more tracks compared to conventional ULM. In some embodiments, using the systems and methods described herein, all vasculature located within the field of view may be imaged with a lower bound of bubbles that move at about 0.5 mm / s. As a result, arterioles, capillaries, and venules may be visualized. Thus, unlike existing techniques, non-invasive classification of vessel structures may be achieved using the systems and methods described herein.

[0121] The systems and methods described herein may therefore be used to unveil spatiotemporal patterns of capillary function, providing new insights into neural pathologies and contributing to the development of early diagnostics in neurodegenerative conditions or new therapeutic interventions. In particular, there may be potential for the systems and methods described herein in the study, treatment, and diagnosis of conditions including, but not limited to, Alzheimer’s, dementia, and stroke. In addition, there may be applications in aging, cognition, and overall brain health monitoring and disease prevention, ensuring equity in healthcare. Compared to existing techniques, the modality proposed herein may also prove to be cost-effective since it requires a reduced number of components (e.g., an ultrasound imaging device 102 and at least one probe 104). Furthermore, the systems and methods described herein may be implemented on already available equipment (e.g., clinical scanners available in a clinic), impacting operation and diagnosis of many neurodegenerative diseases.

[0122] The above description is meant to be exemplary only, and one skilled in the art will recognize that changes may be made to the embodiments described without departing from the scope of the invention disclosed. Still other modifications which fall within the scope of the present invention will be apparent to those skilled in the art, in light of a review of this disclosure.

[0123] Various aspects of the systems and methods described herein may be used alone, in combination, or in a variety of arrangements not specifically discussed in the embodiments described in the foregoing and is therefore not limited in its application to the details and arrangement of components set forth in the foregoing description or illustrated in the drawings. For example, aspects described in one embodiment may be combined in any manner with aspects described in other embodiments. Although particular embodiments have been shown and described, it will be apparent to those skilled in the art that changes, and modifications may be made without departing from this invention in its broader aspects. The scope of the following claims should not be limited by the embodiments set forth in the examples but should be given the broadest reasonable interpretation consistent with the description as a whole.

Claims

WHAT IS CLAIMED IS:

1. A method for vasculature characterization using ultrasound, the method comprising: continuously acquiring ultrasound data from a body comprising a vasculature with one or more ultrasound contrast agents flowing therethrough; filtering the ultrasound data to separate background signals from reflected ultrasound signals generated by the one or more ultrasound contrast agents; tracking, based on the reflected ultrasound signals, an entire trajectory of each of the one or more ultrasound contrast agents through the vasculature; determining, based on the tracking, at least one displacement characteristic of each of the one or more ultrasound contrast agents throughout the entire trajectory over time; and identifying, individually and based on the at least one displacement characteristic, one or more vessels present within the vasculature.

2. The method of claim 1 , wherein filtering the ultrasound data comprises applying a long ensemble singular value decomposition (LE-SVD) clutter filter to the ultrasound data to separate the background signals from the reflected ultrasound signals.

3. The method of claim 2, wherein the LE-SVD clutter filter has an ensemble size greater than 5,000 frames.

4. The method of any one of claims 1 to 3, wherein at least some of the one or more vessels have a diameter below about 100 pm.

5. The method of claim 4, wherein the at least some of the one or more vessels are capillaries having a diameter below about 5 pm.

6. The method of any one of claims 1 to 5, wherein tracking the entire trajectory of the one or more ultrasound contrast agents through the vasculature comprises determining a position of each of the one or more ultrasound contrast agents in each timeframe associated with the reflected ultrasound signals, and generating a plurality of tracks indicative of the position of the one or more ultrasound contrast agents as determined.

7. The method of claim 6, wherein the at least one displacement characteristic is a change in a velocity behavior of each of the one or more ultrasound contrast agents, the change in the velocity behavior determined by filtering the plurality of tracks, spline interpolating the plurality of tracks, and derivating the plurality of tracks with respect to time to estimate a velocity of each of the one or more ultrasound contrast agents.

8. The method of any one of claims 1 to 7, wherein identifying the one or more vessels individually and based on the at least one displacement characteristic comprises determining that a respective one of the one or more ultrasound contrast agents exhibits a U-shaped velocity profile when traversing a respective one of the one or more vessels.

9. The method of any one of claims 1 to 8, further comprising using the at least one displacement characteristic of each of the one or more ultrasound contrast agents to train at least one hidden Markov model to identify the one or more vessels.

10. The method of any one of claims 1 to 9, further comprising detecting an occurrence of at least one vessel stall in the vasculature based on the at least one displacement characteristic of each of the one or more ultrasound contrast agents.11 . The method of any one of claims 1 to 10, further comprising generating a reconstruction of an entire vascular network of the body based on the entire trajectory of the one or more ultrasound contrast agents through the vasculature, and outputting the reconstruction.

12. The method of any one of claims 1 to 11 , further comprising generating, based on the identifying, at least one two-dimensional (2D) super-resolution ultrasound localization microscopy (ULM) image of the vasculature and / or generating at least one pseudo-three- dimensional (3D) super-resolution ULM image of the vasculature, the at least one pseudo- SD super-resolution ULM image providing elevational information about the vasculature, and outputting the at least one 2D super-resolution ULM image and / or the at least one pseudo-3D super-resolution ULM image.

13. The method of any one of claims 1 to 11 , further comprising generating, based on the identifying, at least one Contrast-enhanced ultrasound (CEUS) image of the vasculature, and outputting the at least one CEUS image.

14. A system for vasculature characterization using ultrasound, the system comprising: a processing unit; and a non-transitory memory communicatively coupled to the processing unit and comprising computer-readable program instructions executable by the processing unit for: continuously acquiring ultrasound data from a body comprising a vasculature with one or more ultrasound contrast agents flowing therethrough; filtering the ultrasound data to separate background signals from reflected ultrasound signals generated by the one or more ultrasound contrast agents; tracking, based on the reflected ultrasound signals, an entire trajectory of each of the one or more ultrasound contrast agents through the vasculature; determining, based on the tracking, at least one displacement characteristic of each of the one or more ultrasound contrast agents throughout the entire trajectory over time; and identifying, individually and based on the at least one displacement characteristic, one or more vessels present within the vasculature.

15. The system of claim 14, wherein the instructions are executable by the processing unit for filtering the ultrasound data comprising applying a long ensemble singular value decomposition (LE-SVD) clutter filter to the ultrasound data to separate the background signals from the reflected ultrasound signals.

16. The system of claim 14 or 15, wherein the instructions are executable by the processing unit for tracking the entire trajectory of the one or more ultrasound contrast agents through the vasculature comprising determining a position of each of the one or more ultrasound contrast agents in each timeframe associated with the reflected ultrasound signals, and generating a plurality of tracks indicative of the position of the one or more ultrasound contrast agents as determined.

17. The system of claim 16, wherein the instructions are executable by the processing unit for determining the at least one displacement characteristic comprising a change in avelocity behavior of the one or more ultrasound contrast agents, the change in the velocity behavior determined by filtering the plurality of tracks, spline interpolating the plurality of tracks, and derivating the plurality of tracks with respect to time to estimate a velocity of each of the one or more ultrasound contrast agents.

18. The system of any one of claims 14 to 17, wherein the instructions are executable by the processing unit for identifying the one or more vessels individually and based on the at least one displacement characteristic comprising determining that a respective one of the one or more ultrasound contrast agents exhibits a U-shaped velocity profile when traversing a respective one of the one or more vessels.

19. The system of any one of claims 14 to 18, wherein the instructions are executable by the processing unit for using the at least one displacement characteristic of each of the one or more ultrasound contrast agents to train at least one hidden Markov model to identify the one or more vessels.

20. The system of any one of claims 14 to 19, wherein the instructions are executable by the processing unit for at least one of: generating a reconstruction of an entire vascular network of the body based on the entire trajectory of the one or more ultrasound contrast agents through the vasculature, and outputting the reconstruction; generating at least one two-dimensional (2D) super-resolution ultrasound localization microscopy (ULM) image of the vasculature and / or generating at least one pseudo-three-dimensional (3D) super-resolution ULM image of the vasculature, the at least one pseudo-3D super-resolution ULM image providing elevational information about the vasculature, and outputting the at least one 2D super-resolution ULM image and / or the at least one pseudo-3D super-resolution ULM image; and generating at least one Contrast-enhanced ultrasound (CEUS) image of the vasculature, and outputting the at least one CEUS image.

Images